Intelligent Agents: Agent-Environment Interaction — Beginner-Friendly Notes

1. The Core Loop (learn this diagram by heart)

Environment  →  [Sensors]  →  Agent  →  [Actuators]  →  Environment
      ↑                                                          |
      └────────────────── Percept Sequence ──────────────────────┘

• Agent: anything that perceives and acts (software bot, robot, thermostat).
• Sensors: what agent uses to perceive (camera, microphone, keyboard, temperature sensor).
• Actuators: what agent uses to act (motors, speakers, screen, API call).
• Percept: one snapshot from sensors at one time.
• Percept sequence: entire history of percepts so far — the agent's experience.
• Action: what agent does now, based on the sequence.

Agent function: maps percept sequence → action. Agent program: the actual code that implements that function on hardware.

2. Simple Example — Vacuum Robot

• Sensors: camera, bump sensor, dirt sensor.
• Actuators: wheels, suction motor.
• Percept: [location=A, dirt=yes] → Action: suck.
• Next percept depends on previous action — loop continues.

3. Why This View Matters

• It is universal: human, robot, web crawler, game AI all fit the same loop.
• It turns "intelligence" into engineering: given percepts, choose best action.
• It leads directly to PEAS (next topic) and to agent types (reflex, model-based, etc.).

4. Rational Agent — One Definition

An agent is rational when it chooses the action that maximises its expected performance given its percept sequence and built-in knowledge.

Not "perfect" (cannot know future), just best possible with what it knows.

5. Tiny Contrast

• Ordinary program: input → output (no environment feedback).
• Agent: perceives → acts → environment changes → new percepts → acts again (closed loop).

6. One-Line Summary to Write in Exam

An intelligent agent perceives its environment through sensors, keeps a history of percepts, and acts through actuators to maximise performance — forming a continuous perception-action loop.

7. Quick Self-Check

1. Name sensors and actuators for a self-driving car.
2. What is the difference between a percept and a percept sequence?
3. Why is the loop called "closed"?